Unlocking the key to Chatbot Success

Unlocking the key to Chatbot Success: FORM


Building a successful chatbot requires 3 essential considerations:


  • Function: AI and Data
  • Form: Design and UX
  • Implementation: Internal and External


One can’t do without the others and each element in the process deserves attention from the very beginning.



The Secrets Behind a Great Chatbot User Experience (UX)


A chatbot will never be perceived better than the delivered UX. UX matters. Design matters… that goes for both the visual chat interface and the conversational flow.


So, let’s start from an end:


Designing the chat interface


“It’s all about appearances” would be an overstatement when it comes to chatbots. None the less, when talking UX we need to address the matter.


When it comes to the visual design of the chat interface, a simple layout with no unnecessary distractions is the way to go. Keep it simple and intuitive by drawing attention to the right element at the right time.


It is critical to ensure that the user intuitively knows which action prompts what result. Therefore, bear in mind that distinct interactions require distinct prompts/buttons. Never put the user in a situation where he or she doubts which button/prompt to use. Unpredictable reactions triggers distrust and unclear instruction would lead to just that.


There is a slight dilemma though… Because the simple solution to designing an intuitive interface will often be designing for safe choices. Users tend to be mistrustful of new technologies, and this mistrust translates into safe choices. Using UI components in chat is a great way to keep users on a happy path, but there is a pitfall: safe choices gives the illusion of limitations. It will work perfectly for chatbots built for simple purposes, while limited choices would not work at all for other purposes.


One last important note when it comes to interface design: Chatbots are really not all that different from humans. They are imperfect, and they make mistakes. Acknowledge this and design the interface accordingly by adding an undo or cancel functionality. Always give the user a way out or the ability to go back a step without having to start the conversation all over.


Designing the conversational flow


Each chatbot has its own unique conversational flow. It all depends on the purpose and what the user testing tells you about how your users seek to interact with you chatbot.


But none the less, there are a few golden rules to follow.


First of all, document all possible conversational flows. Unless (or even though) you are building a super-smart AI assistant, you will need some preset scenarios to start with. The AI will be able to learn patterns and scenarios over time, but to begin with you will need to develop conversational flows that account for onboarding, missing inputs, vague or irrelevant requests, request verification and resolution, changing terms or completely abandoning the flow. The more cases the better.


Secondly, guide your user. Interacting with chatbots can be frustrating for people of a certain age or background. Having proper on-boarding and solid user flows helps you engage your users and reduces frustrations. It’s really a matter of providing useful tips and examples on how to use your chatbot and how to ask questions. Always attempt to provide tailored suggestions to keep the user engaged and help him/her succeed with your bot.


Finally, remember that wording matters. On that account there are 4 things worth considering:


  • Clarity over elegance: We are not suggesting that you should always go for the lowest common denominator, but when it comes to designing the conversational flow common wordings are broadly appreciated and understood. Keep it simple and clear, make sure you don’t lose your user on account of preferences.
  • Confirm by asking: Question are a natural way to correct misunderstandings and cuts down on the user annoyance when something goes wrong. A simple way to improve the interaction is to say, “Is it correct that you…?”. Formulating confirmations as questions also allows users to correct their input. Statements, on the other hand, place the blame on the user.
  • The response will follow the question: Be purposeful about the wording of your question, and be especially thoughtful of how you begin your question. A clearly phrased question will get a clear response – a response that actually matches what the user feels he or she was aske
  •  Keep it short: As a rule, aim for no more than 90 characters per message from the chatbot. That is the limit before reaching the “too long – didn’t read” level. The only exception is answers to questions that the user has specifically requested.


General UX tips


Building for an outstanding UX goes beyond the chat interface and the conversational flow. We have gathered a few best practice tips that can help pave the way to chatbot success:


  • Manage expectations: The bot will not understand everything that people say. You need to set the expectations right and tell people upfront what the bot can and cannot understand.
  • Consider dead-ends: You will need to take dead-ends into account and offer an alternative path when the bot knows it’s stuck. It can try to correct itself with the help of the customer. But that will not always be the way out of the dead-end. Therefore, by default always design the chatbot to never leave someone hanging!
  • Consistency: People will only return to use the bot if it’s consistent and accurate. Make sure to build consistently and consider all possibilities within a topic.
  • Build trust: People will learn to trust the chatbot if it’s looking out for them, responsive, transparent, and smart. The biggest hurdle is a smart AI, but it’s not the only trust signal we need to design for. Sentence structure, images and context needs to be taken into account as well.
  • Well-functioning beats impressive: While impressive is great, well-functioning is even better! The user really prefers a well-functioning chatbot that serves its purpose to perfection and lives up to the expectations it sets. There is nothing wrong with designing to impress… as long as you make sure it works perfectly!
  • Think beyond search: Be ambitious (not the same as impressive) and consider thinking beyond traditional search when creating the experience. Admittedly, it may feel like search in its infancy because of the rigidity, but as the AI gets smarter the more the bot can set itself apart and provide what feels like a tailored user experience – something search would never give.
  • Provide a way back: One of the most frequently asked questions during chatbot user tests is: “So… where do I go from here?”. Therefore, make sure that the user knows how to get to the main menu and how to restart the conversation. It could be through a button or a prompt such as: “help” or “start over”.




Looking for the function and implementation articles?


Don’t worry, they are on their way.